Abstract
The utility of studying low-frequency surface weather phenomena with long time series of meteorological data from tropical Pacific islands is demonstrated. The wind stress changes associated with El Niño events in the period 1950–78 are examined at seven locations. Zonal wind stress anomalies at the equator near the date line often exhibit strengthening and subsequent weakening of the trade winds prior to each El Niño, as originally suggested by Wyrtki. An exception is the weak 1963 El Niño, which is preceded by meridional wind stress anomalies at the equator. The strongest zonal and meridional wind stress anomalies, however, occur well after the first occurrence of anomalously warm water off the coast of Peru for each El Niño, in agreement with prior analyses of merchant marine data. Away from the equator, variability of the wind stress anomalies from one El Niño to the next is strong, leading to numerous discrepancies with published profiles of the “mean” El Niño wind changes. Power spectra of wind stress from three island stations are compared with concurrent wind stress spectra computed from merchant marine data. Many disparities are found and can be attributed to (sometimes severe) aliasing in the ship data. Possible aliasing errors in the ship data time series are estimated by randomly subsampling the island data in order to mimic the ship data sampling. Sampling criteria, which depend upon the scientific application, are suggested in order to limit the alias noise in the ship data to acceptable amounts.